Lecture 9.

HYPOTHESIS TESTS
▪ Statistical Hypothesis
▪ Error Types
▪ T-test for Pop. Mean
▪ Z-test for Pop. Proportion
▪ Chi-sq test for Pop. Variance

▪ [1] Chapter 9 + 11
▪ pp. 382 – 432; 488 - 491

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 1

 Lecture 9. HYPOTHESIS TESTS
▪ Case study
• Goal: increasing sales
(current mean: 246)
• New bonus plan for
sales staffs
• Test 40 staffs with new
plan 𝑥ҧ = 270,2 and
𝑠 = 50,7
• What is the suitable pair of hypotheses?
• Comment if 𝐻0 is rejected or not
• If the previous plan is only commission for staffs,
what is your suggestion for the new plan?
PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 2
 Lecture 9. HYPOTHESIS TESTS
▪ Case study
• The manager of a
clothing store and
conduct customer
research
• Observed 122 shopping
customers at the stores,
53 of them are female.
• Is the proportion of male customers higher?
• Observing 200 shopping online customers, 97 of
them are female. The difference between two
group about the female proportion exist or not?
PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 3
 9.1. Statistical Hypothesis
▪ Hypothesis: statement about statistical issues
▪ Testing: test for the “True” or “False” of hypothesis
▪ Example:
• Local authorities test the hypothesis: “the teenager
criminal proportion is lower than last year” (10% is
the data of last year)
• A researcher tests the hypothesis: “the average
interest rate of investors in the stock market is
greater than 2%”
• A migration expert tests the hypothesis: “the
variability of Hanoi income is less than HCM city”
PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 4
 Types of Hypothesis Tests
▪ Two types of hypothesis test
• Parameter test
• Non parameter test
▪ Main researching topic:
• Parameter test on population mean, population
variance and population proportion

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 5

 Pair of hypotheses (Parameter test)
▪ Parameter 𝜃
𝐻0 ∶ Null hypothesis
ቊ
𝐻1 𝑜𝑟 𝐻𝑎 : Alternative hypothesis
▪ There are 3 types
𝐻0 : 𝜃 ≤ 𝜃0
▪ ቊ one-sided, upper tail
𝐻1 : 𝜃 > 𝜃0
𝐻0 : 𝜃 ≥ 𝜃0
▪ ቊ one-sided, lower tail
𝐻1 : 𝜃 < 𝜃0
𝐻0 : 𝜃 = 𝜃0
▪ ቊ two-sided, two tails
𝐻1 : 𝜃 ≠ 𝜃0

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 6

 Example
Ex. Testing the statement that “the average interest rate
of investors in the stock market is greater than 2%”
𝐻0: 𝜇 ≤ 2 : the statement is False
𝐻1: 𝜇 > 2 : the statement is True

Example 9.1. What is pair of hypotheses for the following

statements:
(a) The average income of labors is greater than 2000
(b) The proportion of male customers is less than a half
(c) The variability of price is more than 20 USD2
(d) The dispersion of waiting time is less than 10 minutes
PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 7
 Decision & Errors
▪ Decision: Reject or Accept 𝐻0
▪ Two types of Errors
Type I Error: Reject a True hypothesis
Type II Error : Accept a False hypothesis
Decision 𝐻0 is True 𝐻0 is False
Accept H0 Correct decision Type II error
Prob. = 1 − 𝛼 Prob. = 𝛽
Reject H0 Type I error Correct decision
Prob. = 𝛼 Prob. = 1 − 𝛽
At a given Type I error prob., minimize Type II error prob.
PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 8
 Testing Procedure
▪ Probability for Type I Error: Significant Level, 𝛼
▪ Determine Critical value correspond to 𝛼 and the
statistics distribution
▪ Declare Reject Area for 𝐻0
▪ Compute Statistical value from sample data
▪ Rule:
• Statistical value falls in the Reject Area: Reject 𝑯𝟎
• Stat. value does not fall in the Reject Area: Not
Reject 𝑯𝟎 (Not enough evidences to reject 𝐻0)

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 9

 Testing Procedure
▪ 1st step: Show out Pair of Hypotheses (based on the
statement) and the significance level
▪ 2nd step: Data gathering
▪ 3rd step: Computing statistical value
▪ 4th step: Compare statistical value and critical value
▪ 5th step: Reject the null hypothesis or not and
conclusion on the statement.
▪ There is Probability value (P-value) of the test
▪ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 ≤ 𝛼: reject 𝐻0
▪ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 > 𝛼: not reject 𝐻0
PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 10
 9.2. T-test for Mean
▪ Tests on population mean 𝜇 with a given value 𝜇0
▪ Significant level: 𝛼
▪ Hypotheses pairs
𝐻0 : 𝜇 ≤ 𝜇0 𝐻0 : 𝜇 ≥ 𝜇0 𝐻0 : 𝜇 = 𝜇0
ቊ ቊ ቊ
𝐻1 : 𝜇 > 𝜇0 𝐻1 : 𝜇 < 𝜇0 𝐻1 : 𝜇 ≠ 𝜇0

ҧ 0
𝑥−𝜇
▪ Statistical value: 𝑇 = 𝑛
𝑠
▪ Critical values and Reject Areas are different

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 11

 T-test for Mean

Statistical Hypotheses Critical

Reject H0
value pair value

𝐻0 : 𝜇 ≤ 𝜇0 (𝑛−1) (𝑛−1)
ቊ 𝑡𝛼 𝑇> 𝑡𝛼
𝐻1 : 𝜇 > 𝜇0

𝑥ҧ − 𝜇0 𝐻0 : 𝜇 ≥ 𝜇0 (𝑛−1) (𝑛−1)
𝑇= 𝑛 ቊ𝐻 : 𝜇 < 𝜇 −𝑡𝛼 𝑇< −𝑡𝛼
𝑠 1 0

𝐻0 : 𝜇 = 𝜇0 (𝑛−1)
𝑡𝛼 |𝑇| >
(𝑛−1)
𝑡𝛼
ቊ
𝐻1 : 𝜇 ≠ 𝜇0 2 2

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 12

 Example
Example 9.2. Last year, the average productivity of
workers is 120. Improving workers productivity, a new
production line is applied. A recent survey of 25 workers
has sample mean of 126.2 and sample variance of 225.
▪ Assumed that productivity is Normality,
▪ (a) Test the hypothesis “the current average
productivity has been increased”, at significance level
of 5% and 1%
▪ Sample: 𝑛 = 25; 𝑥ҧ = 126.2; 𝑠 2 = 225 → 𝑠 = 15

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 13

 Example
𝐻0 : 𝜇 ≤ 120
▪ Pair of hypotheses: ቊ
𝐻1 : 𝜇 > 120
126.2−120
▪ Compute 𝑇𝑠𝑡𝑎𝑡 = 25 =
15
▪ At significance level of 5%:
25−1
Critical value: 𝑡0.05 =

▪ At significance level of 1%:

25−1
Critical value: 𝑡0.01 =
▪ Conclusion

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 14

 P-value Approaching
▪ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 of the test is computed by stat. softwares
▪ The rule
• 𝑃 − 𝑣𝑎𝑙𝑢𝑒 ≤ 𝛼 ↔ Reject 𝐻0
• 𝑃 − 𝑣𝑎𝑙𝑢𝑒 > 𝛼 ↔ Not Reject 𝐻0
𝐻0 : 𝜇 ≤ 𝜇0
▪ ቊ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 = 𝑃(𝑇 > 𝑇𝑠𝑡𝑎𝑡 )
𝐻1 : 𝜇 > 𝜇0
𝐻0 : 𝜇 ≥ 𝜇0
▪ ቊ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 = 𝑃(𝑇 < 𝑇𝑠𝑡𝑎𝑡 )
𝐻1 : 𝜇 < 𝜇0
𝐻0 : 𝜇 = 𝜇0
▪ ቊ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 = 2 × 𝑃(𝑇 > 𝑇𝑠𝑡𝑎𝑡 )
𝐻1 : 𝜇 ≠ 𝜇0

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 15

 Example
▪ Example 9.2 (cont.)
▪ (b) At significance level of 5%, test the hypothesis
“the current average productivity equals 130”
▪ (c) Which interval does the 𝑃 − 𝑣𝑎𝑙𝑢𝑒 of the test for
question (b) fall in: (0 – 1%); (1% – 5%); (5% – 10%); >
10%
▪ (d) At significance level of 10%, test the hypothesis
“the current average productivity is less than 135”,
and determine the interval value of 𝑃 − 𝑣𝑎𝑙𝑢𝑒

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 16

 9.3. Z-test for Proportion

Hypotheses Critical
Statistical value Reject H0
pair value

𝐻0 : 𝑝 ≤ 𝑝0
ቊ 𝑧𝛼 𝑍 > 𝑧𝛼
𝐻1 : 𝑝 > 𝑝0
𝑝ҧ − 𝑝0
𝑍= 𝑛 ቊ𝐻0 : 𝑝 ≥ 𝑝0 −𝑧𝛼 𝑍 < −𝑧𝛼
𝑝0 1 − 𝑝0 𝐻1 : 𝑝 < 𝑝0

𝐻0 : 𝑝 = 𝑝0 𝑧𝛼 |𝑍| > 𝑧𝛼
ቊ
𝐻1 : 𝑝 ≠ 𝑝0 2 2

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 17

 P-value Approaching
▪ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 of the test is computed by stat. softwares
▪ The rule
• 𝑃 − 𝑣𝑎𝑙𝑢𝑒 ≤ 𝛼 ↔ Reject 𝐻0
• 𝑃 − 𝑣𝑎𝑙𝑢𝑒 > 𝛼 ↔ Not Reject 𝐻0
𝐻0 : 𝑝 ≤ 𝑝0
▪ ቊ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 = 𝑃(𝑍 > 𝑍𝑠𝑡𝑎𝑡 )
𝐻1 : 𝑝 > 𝑝0
𝐻0 : 𝑝 ≥ 𝑝0
▪ ቊ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 = 𝑃(𝑍 < 𝑍𝑠𝑡𝑎𝑡 )
𝐻1 : 𝑝 < 𝑝0
𝐻0 : 𝑝 = 𝑝0
▪ ቊ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 = 2 × 𝑃(𝑍 > 𝑍𝑠𝑡𝑎𝑡 )
𝐻1 : 𝑝 ≠ 𝑝0

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 18

 Example
Example 9.3. Last month, the proportion of visitors
bought goods is 20%. This month, manager observed 200
visitors, 52 of them bought at least one item.
(a) With significant level at 5%, test the hypothesis “the
proportion of customers has increased”
(b) Find out 𝑃 − 𝑣𝑎𝑙𝑢𝑒 approximately for the test
(c) At significant level of 1%, test the hypothesis that
Proportion is 30%, and determine the 𝑃 − 𝑣𝑎𝑙𝑢𝑒

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 19

 9.3. Chisq-test for Variance
Statistical value Hypotheses pair Reject H0

𝐻0 : 𝜎 2 ≤ 𝜎02 2 𝑛−1
൝ 𝜒2 > 𝜒𝛼
𝐻1 : 𝜎 2 > 𝜎02
𝐻0 : 𝜎 2 ≥ 𝜎02 2 𝑛−1
𝑠 2 (𝑛 − 1) ൝ 𝜒 2 < 𝜒1−𝛼
𝐻1 : 𝜎 2 < 𝜎02
𝜒2 =
𝜎02
2 𝑛−1
𝜒 2 > 𝜒𝛼
𝐻0 : 𝜎 2 = 𝜎02 2
൝
𝐻1 : 𝜎 2 ≠ 𝜎02 𝜒2 <
2 𝑛−1
𝜒 𝛼
1− 2

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 20

 P-value Approaching
▪ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 of the test is computed by stat. softwares
▪ The rule
• 𝑃 − 𝑣𝑎𝑙𝑢𝑒 ≤ 𝛼 ↔ Reject 𝐻0
• 𝑃 − 𝑣𝑎𝑙𝑢𝑒 > 𝛼 ↔ Not Reject 𝐻0
𝐻0 : 𝜎 2 ≤ 𝜎02 2
▪ ൝ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 = 𝑃(𝜒 2 > 𝜒𝑠𝑡𝑎𝑡 )
𝐻1 : 𝜎 2 > 𝜎02
𝐻0 : 𝜎 2 ≥ 𝜎02 2
▪ ൝ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 = 𝑃(𝜒 2 < 𝜒𝑠𝑡𝑎𝑡 )
𝐻1 : 𝜎 2 < 𝜎02
𝐻0 : 𝜎 2 = 𝜎02 2 < 𝜒2 2 > 𝜒2 )
▪ ൝ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 = 2 × 𝑚𝑖𝑛 𝑃 𝜒 𝑠𝑡𝑎𝑡 , 𝑃(𝜒 𝑠𝑡𝑎𝑡
𝐻1 : 𝜎 2 ≠ 𝜎02

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 21

 Example
Example 9.4. The required limit for the variability of
fruits’ weight is 4g2. For sample of 20 fruits, sample
variance is 5g2.
a) With the significant level of 5%, test the hypothesis
“the variability of fruits’ weight exceeds the required
limit” (assumed that the weight has normal
distribution).
b) Test the hypothesis that Population Standard
deviation is less than 4g

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 22

 Key Concepts
▪ Hypotheses pair
▪ Types of Error
▪ Significant level
▪ Critical value & Statistical value
▪ Reject and Not reject H0
▪ 𝑃 − 𝑣𝑎𝑙𝑢𝑒 of the test
▪ T-test, Z-test, Chisq-test

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 23

 Exercise
[1] Chapter 9, 11
▪ (387) 1, 2, (389) 5, 7
▪ (402) 10, 15, 17
▪ (408) 23, 25, 27, 29, 34
▪ (414) 36, 38, 40, 43
▪ (491) 4, 11
▪ Case Problem 1

PROBABILITY & STATISTICS – Nguyen Hai Duong – NEU – www.mfe.edu.vn/nguyenhaiduong 24